i'm helping a client design a house that'll be built using concrete and rebar, which may significantly impact his WiFi coverage. Rather than put an AP in every room (and deal with the problems of seamless roaming), does anyone make a "leaky coax" that I can string around the house? I've seen this used in tunnels and subways for radio gear...
Not for 2.4GHz. The highest frequency in use is 900MHz and that's with fairly high power radios. The problem is that it's still coax cable whether it leaks or not and coax cable at 2.4GHz is very lossy. By the time you would snake it around the house, most of your signal will be absorbed by the coax, not radiated.
Since this construction is new, it would be easy enough to install signalling conduits in the walls, and snake whatever cables are required for a wired LAN later. With concrete construction, the last thing the owner needs is a jack hammer remodel. It's not just the LAN but CATV coax, rotator cable, intercom, telco, fiber, hi-fi, satellite dish, free to air, ham radio, GPS time sync, and whatever else I forgot that will be a problem with wiring. Put in BIG conduit, especially to the roof, leave lots of room for expansion, leave a non-rotting pull string, and forget about wireless for now. Make sure everything comes to one place (19" rack, communications closet, or structured wiring panel) with lots of room for punch down blocks (type
110), LAN boxes, and standby power UPS.
If the owner insists on wireless in each room, use the CAT5 LAN wiring in the conduit to act as a backhaul for multiple access points. Methinks you'll find the $50-$70 cost of one access point per room to be much less than any exotic RF solutions.
Seamless, as in roaming from room to room without dropping connections? No problem.
The "problem" with roaming wireless is in roaming between access points that each have different routes to the internet via different ISP's. Each access point would assign a different IP address, gateway, and DNS server as the client radio moves among the access points. At this time, there's no common standard for transfering an IP address between access points, but it's being worked over and will be part of some future 802.11(something) standard.
However, that's NOT a problem with a home network of multiple access points. Each access point has the exact same route to the internet and delivers the exact same IP addresses to the client from a single DHCP server. Setup all access points with the same SSID and WEP/WPA key, select channels for the minimum interference between adjacent access points, and a wireless client will roam merrily among them. There will be a slight delay as the client switches between AP's, but the connection will resume momentarily. It's exactly the same as if you unplugged a wired computah from an ethernet switch, and moved the cable to another port on the same (or nearby) switch.
Some Windoze clients are better at roaming than others. XP SP2 does a decent job of it. I really haven't played with this much. If you really want seamless and like spending money, try one of the "wireless switch" vendors, where the radios are literally brain dead, and all the action happens in the central switch.
and Cisco also have wireless switch products. However, methinks this is overkill for a home network. It can be done with generic access points, but it would be a good idea to buy two and test it first to avoid suprises. I'm guessing(tm) that you'll have more problems with the various wireless clients not wanting to let go of a connection from an AP with a marginal signal, than with any problems inherent in the AP's.
Oh, yes, we're planning lots of conduit, but one of the requirements is seamless wireless networking, and AFAICT that's not possible with the current state of the art in residential-class APs (unless there's something I'm missing).
Actually "Leaky Coax" is available for 2.4Ghz, but it is quite costly.. I have worked with many different distributed antenna systems for 2.4.. Many Installers of Malls, Hotels, Train Tunnels, i.e. Sprint etc.. use DAS instead of gobs of AP's...
Andrew makes a great one for 2.4 that I have used in a tunnel in Europe.. It is a combination of their RADIAX and HELIAX cables:
Very cool stuff that works, but probably way way too costly for your home.. Besides 5/8" hardline is not easy to pull anyway :)
Does it work with 802.11b/g? When I tried some lab experiments with multiple antennas on a coax run, I found that multipath was just killing the data with delay spreading and generating intersymbol interference. I got great signal signal strength, thruput just stunk.
We didn't use leaky coax as we didn't want to lose signal by radiating inside walls, risers, ducts, etc. The system was a number of fairly high impedance radiators, located on multiples of a half wave electrical along the coax. The end was terminated to eliminate reflections. The match was tolerable (-15dB return loss) mostly thanks to cable loss. A typical large room had 4 radiators.
It worked well if I had only one radiator per room, but would screwup badly with 4 due to multipath. I never tried leaky coax because I assumed it would have the same problem. However, this was before
802.11g, where OFDM may have helped deal with the reflections.
Leaky Coax works well with 11b, and actually better with 11g as OFDM is more forgiving. Multi-path has not been an issue with our installs as we only run 1 radiator in an area... Even with multiple radiators, each antenna is only x-mitting and rcving at the same instance (TDD) as it is hooked to the same Radio.. Multi-Path from the perspective of the client has not happened... I can only guess why.. perhaps because the area from the cable that is in the radiation zone is small, low DB after a reflection and phase shift, the difference in gain is great hence no problem??? We have actually loaded Radio's farther down the line with Duplexors made by Motorola (but I am not sure how they take care of issues.. already a done design) This was done on a Metal Vessel..
3 Radios per cable... 1,6,11
The only problem with the design of these systems and 802.11 is CSMA... Hidden node can become an issue for obvious reasons.. so RTS/CTS helps... I have found great results in limited clients, and good results for internet access which can take collision reduced bandwidth with higher number of users...
For a lower loss for 2.4Ghz but a super high price, there is Leaky Waveguide... but will suffer the same problems.
The stuff is just to expensive for me to experiment with for more qualified data :)